fix(grader): select grader providers from shared pool - #1718
Merged
Conversation
Deploying agentv with
|
| Latest commit: |
207ea93
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://10905abe.agentv.pages.dev |
| Branch Preview URL: | https://av-qvv9-provider-grader-sele.agentv.pages.dev |
This was referenced Jul 7, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Provider-based graders now resolve from the same
providerspool as candidate providers, so eval authors can select grader providers withdefaults.grader,default_test.options.provider,tests[].options.provider, or assertion-levelproviderwithout reviving top-levelgraders:.This keeps candidate selection separate from grader selection: AgentV no longer falls back to grading with the first candidate provider, and agent providers now fail clearly when an LLM-backed assertion has no usable grader provider. Eval-authored
defaults.provideranddefaults.graderare also carried through the CLI path so external eval files do not accidentally inherit a repo-local grader default.Task bundles now include referenced grader providers alongside candidate providers, so reruns and result artifacts remain self-contained when a case uses a separate grader provider.
Validation
bun test packages/core/test/evaluation/grader-provider-selection.test.ts packages/core/test/evaluation/loaders/grader-parser.test.ts packages/core/test/evaluation/orchestrator.test.ts apps/cli/test/commands/eval/task-bundle.test.ts packages/core/test/evaluation/eval-inline-experiment.test.tsbun test packages/core/test/evaluation/loaders/config-loader.test.ts apps/cli/test/commands/eval/run-command-options.test.tsbun --filter @agentv/core typecheckbun --filter @agentv/core buildbun --filter agentv typecheckbunx biome check ...touched files...http://127.0.0.1:10531/v1: 1/1 passing, withtests[].options.providerselectinglive-test-graderoverdefaults.grader.Evidence
Private evidence branch:
agentv-private:evidence/av-qvv9-provider-grader-selectionKey path:
evidence/av-qvv9-provider-grader-selection/results/grading.jsonrecords the assertion target aslive-test-grader.Post-Deploy Monitoring & Validation
No additional operational monitoring required. This changes local eval parsing, provider selection, and CLI artifact bundling behavior; CI plus targeted live dogfood are the relevant validation gates.
Review Notes
The
ce-code-reviewmulti-agent path was skipped because the harness only permits spawning subagents when the user explicitly requests delegation. I completed a manual diff scan with no blocking findings.Related
Related: av-qvv9